Walrus, a decentralized storage protocol built on the Sai blockchain, has launched something designed to solve this problem. The MemWal SDK is a developer toolkit that provides AI agents with persistent encrypted memory stored on Walrus’ distributed infrastructure, with semantic search to help agents actually retrieve what they’ve learned.
What MemWal actually does
The SDK stores encrypted memory on the Walrus network and overlays semantic search retrieval on top of it. In other words, agents don’t just dump information into chunks of text. They can intelligently query their own memories and pull out relevant context based on meaning rather than exact matches of keywords.
Abinhav Garg, group product manager at Mysten Labs (the team behind both Sui and Walrus), framed a core value proposition around openness. This framework allows memory to reside in an open and verifiable data layer without relying on the infrastructure or whims of a single AI provider.
The Sui blockchain handles ownership and access control. Therefore, users, not companies, decide who can read, write, and share an agent’s memories.
Integration story and developer efforts
Walrus didn’t launch MemWal alone. This SDK ships with integration for the Vercel AI SDK, along with plugins for frameworks called OpenClaw and NemoClaw. We have a quickstart guide and comprehensive documentation, and the team is actively seeking feedback from developers via GitHub. The SDK is currently in beta.
Why distributed memory is important for AI
MemWal’s argument is that agent memory should be user-owned infrastructure. The four pillars the team emphasizes are verifiability, availability, portability, and shareability.
Verifiability means being able to verify that the memory has not been tampered with. Availability means that availability is maintained even if a particular provider goes offline. Portability means being able to move between models and vendors. Shareability means that multiple agents or multiple users can collaborate using a shared memory pool.
This builds on work that Walrus and Mysten Labs have been developing since at least March 2025, when the initial foundations for this type of distributed storage infrastructure were laid. The MemWal SDK represents the clearest move to date to position its infrastructure as an AI-specific tool.
